The Innovation of Ceramic Interpreter Cultivation Mode under the Perspective of Intercultural Communication

The significance of ceramic arts and crafts education for the preservation of traditional culture and skills is self-evident. Ceramic art occupies a pivotal position in human civilization, carrying a large amount of historical information and cultural value. These special cultural heritages can be passed on from generation to generation through education, so that the national culture always maintains continuity and vitality.

In today’s modern production methods continue to rise, handmade ceramic art is facing a crisis for mechanized production replacement. The planned and step-by-step development of teaching programs by educational institutions ensures that these skills are not lost over the years, but rather are more widely circulated and practiced.

Ceramic arts and crafts education not only educates students on traditional techniques, but also encourages them to explore and innovate artistically by stimulating their creativity and aesthetic ability beyond the traditional framework.

Through continuous experimentation and practice, ceramic arts and crafts education fosters students’ independence of thought and encourages them to reinterpret traditional materials, techniques, and forms. This approach has helped students to develop a unique design language, which has led to the emergence of new ideas and new works in the ceramic art world, and has injected fresh vitality and inspiration into the entire art world.

While education trains ceramic designers with artistic cultivation, it also provides the ceramic industry with professionals who understand the market demand, master modern manufacturing technology, and the concept of environmental protection. These people can incorporate new technologies and materials into the production process to optimize the process, reduce energy consumption, and reduce waste of raw materials.

Ceramic arts and crafts education stimulates sustainable design thinking and promotes product design that considers the impact of the environment, which has a positive impact on the entire life cycle of ceramic products. Applying eco-friendly materials and clean energy to education and teaching enables the ceramics industry to take on social responsibility and reduce the burden on the environment, while also seeking economic benefits.

1)

It meets the needs of compound and applied foreign language talents training. With the further deepening of reform and opening up, the market has put forward higher requirements for English talents. The society requires English talents to have solid professional knowledge, high information literacy and the ability to communicate in English in daily life and professional fields. That is to say, the talents cultivated by English majors in the future should be the composite management talents who are proficient in English, have the ability of cross-cultural communication and have specialties. Internationalized economic and trade talents who have knowledge of foreign affairs and are good at business negotiation. With a multi-talented export-oriented composite talent. Ceramic interpreter training is in line with the needs of the times and fully embodies the composite, applied foreign language personnel training requirements.

In the context of AI empowerment, the curriculum design and implementation of interpreting teaching in comprehensive universities need to incorporate new technologies and new teaching methods to adapt to the ever-changing interpreting needs and students’ learning needs. 1)

Establishing an interpreting corpus

Interpreting teaching resources corpus is a fundamental project of interpreting education informatization, which is of great significance in enriching the content of interpreting teaching, meeting the demand for personalized training and optimizing the path of interpreting teaching. With the standardization and systematization of interpreter training, the corpus of interpreting teaching resources has become an important part of interpreting teaching [19].

Interpreting teaching resources corpus is the basic project of interpreting education informatization, which is of great significance in enriching the content of interpreting teaching, meeting the demand for personalized training, and optimizing the path of interpreting teaching. With the standardization and systematization of interpreter training, the corpus of interpreting teaching resources has become an important part of interpreter training.

Figure 1.

The ceramic interpretation teaching mode of information technology

1)

Audio Retrieval Technology

This paper proposes an audio retrieval algorithm based on ASF features, and the main contribution to the algorithm is reflected in the proposed format normalization in preprocessing and the parameter matching adjustment during retrieval. The format normalization in preprocessing can unify the processing of audio with different formats and sampling rates. By adjusting the parameters of the LSH algorithm, a faster retrieval speed can be obtained.

The process of audio retrieval involves indexing all the library sample audio after feature extraction. Then the index and the feature file are saved together to form the audio sample database. When a query audio input, in accordance with the same algorithm for feature extraction. Then, through the index and part of the database library, audio feature files are compared, and finally, through scoring, we return to the most similar audio files.

In terms of feature selection, this paper relies on ASF features. Meanwhile, the preprocessing of the initial audio is improved, and the code rate and sampling rate are unified. In terms of retrieval, a locally sensitive hash algorithm is used to construct the index structure. This form of indexing is convenient and fast, but it can have problems such as internal duplication. In this paper, we will explore the possibility of finding reasonable parameters through continuous debugging. Finally, in terms of engineering practice, OpenMP technology is utilized to add multithreading for feature extraction and retrieval in order to enhance time performance.

Audio feature extraction is divided into three stages: audio data preprocessing, ASF feature calculation, and post-processing.

where N is the number of sample points in a frame, i ∈ [0, N]. In this paper, since the sampling rate has been normalized, N = 512 has been determined when setting the length of each audio frame to 64ms.

Since the ASF feature is an audio frequency domain characterization, it is necessary to transfer each frame of audio data from the time domain to the frequency domain after the frame-splitting process. Commonly used time domain to frequency domain methods include discrete Fourier transform, discrete cosine transform and Haar transform. In this paper, the discrete Fourier transform is chosen. The basic formula of discrete Fourier transform is: (2)X(k)=DFT[x(n)]=∑n=0N−1x(n)e[−j*2π*n*k/N]$$X(k) = DFT[x(n)] = \sum\limits_{n = 0}^{N - 1} x (n){e^{[ - j*2\pi *n*k/N]}}$$

After obtaining the audio frequency domain segments, the values are normalized. The normalization formula is as follows: (3)X1(i)=X(i)−μσ$${X_1}(i) = \frac{{X(i) - \mu }}{\sigma }$$ (4)X2(i)=X1(i)−min[X1(i)]max[X1(i)]−min[X1(i)]$${X_2}(i) = \frac{{{X_1}(i) - \min [{X_1}(i)]}}{{\max [{X_1}(i)] - \min [{X_1}(i)]}}$$

Where X(i) denotes the ind data point in each audio frame, μ is the mean value of the frame, and σ is the variance of the frame. min and max are the maximum and minimum values in X₁(i). After normalization, the range of values of the data points in each frame will be unified as [0, 1].

The values of data points are normalized only to eliminate the effect of signal strength. In order to reduce the effect of other white noise or Gaussian noise, other filters need to be utilized. In this paper, a filter bank consisting of a median filter, a Gaussian filter and a Hamming filter is chosen to filter the normalized audio data. Only 3 sets of cascade filters are used in this paper. The human ear is a complex audio processing system consisting of three parts: the outer ear, the middle ear and the inner ear.

Outer ear: (5)Aih(fklt)=−2.184(f1000)−0.8+6.5e−0.6(f1000−3.3)*−0.001(f1000)36$${A_{ih}}({f_{klt}}) = - 2.184{(\frac{f}{{1000}})^{ - 0.8}} + 6.5{e^{ - 0.6{{(\frac{f}{{1000}} - 3.3)}^*}}} - 0.001{(\frac{f}{{1000}})^{36}}$$

Middle ear: (6)W′(f)=10Aμ(f)/20$${W^\prime }(f) = {10^{{A_\mu }(f)/20}}$$

Using the human ear response filter described above, the human II. Sensitive regions of the audio data can be retained, and the regions that are not sensitive to the human ear can be filtered.

In addition to the consideration of Gaussian white noise, it is also necessary to consider the interference brought about by the change of audio rate, etc. After research, it is found that the inverse filter can be more effective to each other this kind of interference, and its mathematical expression is: (7)yn=xn+∑i=1Maixn−i$${y_n} = {x_n} + \sum\limits_{i = 1}^M {{a_i}} {x_{n - i}}$$

where x_n is the current point data value, ∑iMaixn−1$$\sum\limits_i^M {{a_i}} {x_{n - 1}}$$, is the weighted sum of all points before x_n, a_i is the weights, and M is the filter order. In order to cope with the new noise again, a pre-emphasis window and a Hamming filter bank are used to finalize the data to serve the purpose of reducing the new noise and re-smoothing. (2)

Calculation of ASF features

The ASF feature responds to the smoothness of the audio data spectrum and performs well in many audio related applications. Its feature calculation formula is as follows: (8)F1=(∏n=0N−1x(n))1N1N∑n=0N−1x(n)$${F_1} = \frac{{{{(\prod\limits_{n = 0}^{N - 1} x (n))}^{\frac{1}{N}}}}}{{\frac{1}{N}\sum\limits_{n = 0}^{N - 1} x (n)}}$$

From the formula, the ASF frequency domain segmentation feature values take the range of [0, 1].

Before calculating the ASF features, each audio frame is segmented in the frequency domain.250Hz to 2000Hz is the most sensitive frequency range for the human ear. Therefore, based on this range, the audio data is partitioned into a number of subband segments in the logarithmic frequency domain space. The subbands do not overlap each other and the number of subbands conforms to Eq: (9)bandNum=log2(hiFre/loFre)octaveResolution$$bandNum = \frac{{{{\log }_2}(hiFre/loFre)}}{{octave\:\operatorname{Re} \:solution}}$$

where hiFre and loFre denote the upper and lower limits of the frequency domain. octave:Re solution is the relaxation variable, whose value range is generally an integer power of 2 in the interval [1/16, 8].

ASF_i = [F₁, F₂, F₃, ..., F_N], where i denotes the frame number, and N is the number of feature points per frame, that is, the number of subbands computed. In this paper, an audio signature will be used to compress the cascaded audio features. Namely: (10)ASF=[ F0.0 ... F0.N−1 ... ... ... FM−1.0 ... FM−1.N−1]$$ASF = \left[ {\begin{array}{c} {{F_{0.0}}}&{...}&{{F_{0.N - 1}}} \\ {...}&{...}&{...} \\ {{F_{M - 1.0}}}&{...}&{{F_{M - 1.N - 1}}} \end{array}} \right]$$

The cascaded audio features can be represented in the form of a matrix as above. The signature compression method does not reduce the differentiation and robustness of the features. The method first sets the compression factor df, in this paper, df = 24. Then the number of rows of the matrix after signature compression will become b, b = [M/df]. Then the above matrix is finally compressed into a b × N-dimensional matrix. The compression formula is as follows: (11)AS(k,j)=1df*∑i=1dfASFi.j,0≤j≤N−1,1≤k≤b$$AS(k,j) = \frac{1}{{df}}*\sum\limits_{i = 1}^{df} A S{F_{i.j}},0 \leq j \leq N - 1,1 \leq k \leq b$$

where k is the row index of the compressed matrix, and the final b × N-dimensional matrix is expanded and concatenated into a 1 × bN-dimensional vector.

In this chapter, index structure construction, a key technique for music similarity retrieval, is investigated for the above problem. A representative index structure construction algorithm, Filter-Refine (FR), is investigated and used as the baseline algorithm for the study.

The filter-refinement (FR) index structure construction method maps the Gaussian features of the G1 measure to the Euclidean space by the fast projection algorithm (FM). Following that, the FR index structure is constructed by indexing the musical similarity measure with Euclidean vectors. The G1-FR music similarity retrieval algorithm is proposed to greatly simplify the computational complexity of finding nearest neighbor music objects. The algorithm improves the retrieval speed by 10 to 40 times at the expense of a small loss of retrieval accuracy when applied to a million-volume dataset.

Assuming that each object (music) in the dataset corresponds to a point in the original distance space, the FM algorithm passes the distance function or proximity function in the original distance space d(o_i, o_j). By certain calculations, each point is projected into the k-dimensional space that is orthogonal to each other, and the data is downscaled or vectorized according to the different objectives.

Suppose there is a triangle ABC in two-dimensional space, A, B, C are the vertices, where vertex C represents the object point a, points B and C represent the two axes x_{j, 1} and x_{j, 2} respectively that determine the straight line AB, the AC side represents d(a, x_{j, 1}), and the BC side represents d(a, x_{j, 2}). The projection of C on the straight line AB is AD, and Eq. (12) can be deduced from the cosine theorem: (12)BC2=AC2+AB2−2⋅AD⋅AB$$B{C^2} = A{C^2} + A{B^2} - 2 \cdot AD \cdot AB$$

It can be deduced from Eq. (12) that the projection AD of any object a on a straight line in space can be obtained from Eq. (13) when the points x_{j, 1} and x_{j, 2} of any straight line axis are known: (13)Fj(a)=d(a,xj,1)2+d(xj,1,xj,2)2−d(a,xj,2)22d(xj,1,xj,2)$${F_j}(a) = \frac{{d{{(a,{x_{j,1}})}^2} + d{{({x_{j,1}},{x_{j,2}})}^2} - d{{(a,{x_{j,2}})}^2}}}{{2d({x_{j,1}},{x_{j,2}})}}$$

The above is the mapping method of points in two-dimensional space, which is generalized to multi-dimensional space.The schematic diagram of FM multi-dimensional space mapping is shown in Fig. 2.

H is a hyperplane in n − 1-dimensional space perpendicular to O_aO_b. If O_i, O_j is two object points in n-dimensional space, their projections mapped to the H hyperplane are O_m and O_n, respectively, and the distances d(O_m, O_n) between them can be calculated by the distances d(O_i, O_j) in n-dimensional space, as shown in Eq. (14): (14)d2(Om,On)=d2(Oi,Oj)−(xi−xj)2$${d^2}({O_m},{O_n}) = {d^2}({O_i},{O_j}) - {({x_i} - {x_j})^2}$$

Similarly, the mapping of object points in the original space to any k-dimensional space can be accomplished in k iteration. The specific steps of the iteration are detailed below:

STEP1/ Randomly select a piece of music x_r in the music dataset.

STEP2/Calculate the distance between other music and x_r according to the original distance calculation formula, and take the music with the largest distance as the first axis point x_{j, 1}.

STEP3/ Calculate the music with the largest distance x_{j, 1} according to the original distance calculation formula and take the music with the largest distance as the second axis point x_{j, 2}.

When the 2k axis points of the k-dimensional Euclidean space are selected, the k-dimensional Euclidean space vector representation of any music a in the dataset can be calculated by equation (14). If there are N object points in the dataset, the input of the FM algorithm is the N × N-dimensional distance matrix of the original space, and the output is the N k-dimensional vectors of the specified dimension (k) Euclidean space.

Figure 2.

FM A schematic representation of the multidimensional spatial mapping

After mapping the Gaussian features in the G1 music similarity measure to high-dimensional vectors in Euclidean space using the FM algorithm, the G1-FR algorithm first filters the music using the mapped Euclidean vectors as indexes. That is, given the music object to be retrieved. First, calculate the Euclidean distance between the mapped Euclidean vector of this object and the mapped high-dimensional vectors of other objects and return a certain number of objects as candidates. Then, the K-L scatter is computed in the candidate subset using the original proximity degree to refine the retrieval results. Since the complexity of a single scatter calculation is much higher than a single Euclidean distance calculation, the use of Euclidean distance to pre-filter the retrieval results can substantially improve the retrieval speed compared to the use of scatter to exhaustively match in the dataset.

The G1-FR music similarity retrieval algorithm is divided into two phases: an offline indexing phase and an online retrieval phase. 2)

Ceramic Professional Interpreter Corpus

The overall framework of the Ceramic Interpretation Corpus is shown in Figure 3.

The construction of the platform adopts a “five-layer” design, namely, data collection, data processing, data function, data management and data exchange, which support each other interactively and coordinate with each other. The first layer is responsible for screening and collecting the original corpus of interpretation, which forms the original material basis of the platform database. The second layer is responsible for the secondary processing of the raw corpus to make it become the main object and material for later application and interpretation, which involves the annotation type, level and transcription procedure, and is the key technical guarantee of the whole project. The third layer is responsible for designing the main functions, coordination mechanism, dynamic system, and management model of the entire corpus platform, which is the core of the actual application of the entire platform. The fourth and fifth layers are responsible for developing and managing the corpus linking and sharing technology, which is the final manifestation of the cross-border application of the entire platform.

Figure 3.

The overall framework of ceramic interpretation

In the midst of this research, three types of forms, namely tests, questionnaires and interviews, will be utilized to address the research questions. 1)

Testing: this experiment contains pre-test and post-test. At the beginning of the experiment, the ceramic interpreting level of the two classes will be tested. To comprehend the present learning situation of the two classes and ensure that there are no significant differences in the level of ceramic interpreting between the two classes. The data was entered and analyzed by SPSS software to ensure the reliability and validity of the study.

After half a semester of training in different teaching styles, the professional level of interpreting ceramics in experimental and control classes was tested again. The post-test was designed to evaluate the professional ceramic interpreting performance of the students in the experimental class after a period of experimentation with teaching methods assisted by the interpreting corpus. The data was also processed and analyzed by SPSS software.

The whole research process was divided into three phases of experiment, pre-test teaching experiment process, questionnaire, personal interview and post-test experiment.

Before conducting the empirical study, the pre-test was administered to the experimental and control classes on March 1, 2024, and the scores were entered and analyzed through SPSS software after the pre-test was corrected. The aim was to ensure that the students in both classes had comparable levels of ceramic interpreting learning. On the assumption that the experimental and control classes had comparable levels of ceramic interpretation, the two phases of the study continued.

The experiment began on March 4, 2024, lasted for eight weeks, and ended on April 27, 2024. In the experiment, the control class was taught according to the traditional ceramic interpreting teaching method, i.e., the teacher explains, gives examples, translates, and the students practice. The experimental class was taught ceramic interpreting with a 3-part online and offline blended instruction based on an interpreting corpus.

At the end of the experiment, questionnaires were distributed and personal interviews were conducted in the experimental class. Random interviews were conducted with professional teachers of different titles and teaching grades. And on April 27, 2024, a posttest was administered to the experimental and control classes. The post-test scores were also entered and analyzed through SPSS software.

Here, the pre- and post-test ceramic interpretation scores of the control and experimental classes were transferred and a paired-samples t-test was performed on these two sets of data using SPSS. The pre-test and post-test scores are shown in Table 1. The ceramic interpreting scores of the experimental and control classes before the test were kept around 65, which was at the level of basic pass. After the test, the ceramic interpreting scores of the students in the experimental class were able to achieve about 80 points, indicating that there is a significant difference between the ceramic interpreting scores of the students in the experimental class before and after the test.

The paired samples t-test for pre and post-test scores within the class is shown in Table 2. In the paired test, the Sig. (two-tailed) value of the pre- and post-test scores of the experimental class is 0.000, which is less than 0.01. The Sig. (two-tailed) value of the pre- and post-test scores of the control class is 0.032, which is less than 0.05 and greater than 0.01. This indicates that there is a significant difference in the scores of the students of the experimental class and the control class before and after the experiment. However, the improvement in the performance of the students in the experimental class was greater and more significant than that of the students in the control class. It has been proven that the interpreting corpus-assisted teaching model is effective and has resulted in a significant improvement in the students’ performance in ceramic interpreting.

In this study, a questionnaire was distributed once to 60 students in an experimental class after an 8-week experiment of teaching ceramic interpreting with the aid of an interpreting corpus. A total of 60 questionnaires were distributed and 60 were returned.

The results of the post-experimental questionnaire for the experimental class are shown in Table 3. The post-experimental questionnaire was investigated from the following perspectives.

(Questions 1-3) Whether the students felt that the blended teaching mode of ceramic interpreting assisted by the interpreting corpus improved their self-reading and subsequent writing abilities.

(Questions 4-5) Whether students felt that the blended mode of teaching ceramic interpreting with the aid of an interpreting corpus improved their attitudes towards interpreting in ceramics and their ability to learn on their own.

(Questions 6-8) How much do the students recognize the blended teaching model of ceramic interpreting with the aid of the interpreting corpus, and whether they will continue to use the interpreting corpus to assist their interpreting practice in their future studies.

In questions 1-3 about the impact of the blended teaching mode of ceramic interpreting assisted by the interpreting corpus on students’ ceramic interpreting ability, 78% of the students thought that their ceramic interpreting performance improved after using the interpreting corpus. 70% of the students thought that their thoughts were clearer and they could correctly understand the meaning of the text after using the interpreting corpus, and 72% of the students thought that the interpreting corpus made ceramic interpreting more professional.

From the analysis of the above data, it can be seen that after the experiment, most of the students’ ceramic interpreting ability has been improved, and the blended teaching mode of ceramic interpreting assisted by the interpreting corpus can help to increase the students’ interest in ceramic professional interpreting, help the students to clarify the ideas of interpreting, and make the language of interpreting more authentic and authentic, and enhance the students’ independent learning ability. Most of the students recognize this learning mode and hope to use it in their future learning.

1)

Analysis of students’ overall demand for vocational ability

The data on the mean value of students’ demand for vocational competence is shown in Table 4. The mean values of students’ demand for professional competence, inter-professional competence, methodological competence, and social competence are 4.27, 4.325, 4.145, and 4.0175, respectively, indicating that students’ demand for all four types of competence is relatively high, especially for professional competence and inter-professional competence.

In terms of specific indicators, students majoring in Ceramic Interpreting have a demand for all the other occupational competencies above 4.0, except for the mean value of the demand for innovative ability, work attitude, and teamwork ability, which is below 4.0. The top three occupational ability demands with high demand rate are interpreting ability, problem solving ability demand and English application ability demand respectively. It can be seen that ceramic interpreting majors have a high demand for vocational competence and are generally eager to improve their various vocational competence.

According to the survey, the overall evaluation of students on the effectiveness of the vocational competence development they received is high, in terms of the mean value, the mean value of students’ evaluation of professional competence, inter-professional competence, methodological competence and social competence are 3.773, 3.847, 3.810 and 3.905 respectively, all of which are moderately high. The mean values of students’ evaluation of vocational competence development are shown in Figure 4.The mean values of 16 competence development evaluations are all above 3.5, and the recognition of C8 language organization and comprehension, C9 learning ability needs, and C15 environmental resilience needs are in the top three, with 3.991, 3.967, and 3.997, respectively.

Figure 4.

The student’s professional ability to cultivate the mean